Name: Facebook's Plan to Contain Hate Speech
Uploaded: 2018-05-21
Duration: 5 min 25 s
Description: Will hiring thousands of moderators actually work?

[ MUSIC ] The embattled social media giant just announcing changes. Yeah, flexing their muscles, aiming to take action against- [ UNKNOWN ] in a pool of blood on the ground, all of it broadcast live on Facebook. Facebook and other social media sites are being criticized for not doing enough to stop- Governments are turning up the heat on social media companies by You have created these [ UNKNOWN ] and now they are being misused. It's not enough to just build powerful tools. We need to make sure that they are used for good, and we will. Can you start out just telling us what your role is, here at Facebook? I manage the content standards at Facebook. That's basically the rules for what you can and cannot post. When you come to the site. I have joined the company about six years ago and we already had a team in place at that time thinking about what the community standards or content standards should be. And those content standards have always been public but they have changed over time as the issues in the real world had Change. We've also updated those standards and given people more and more details about how we think about issues over time. These practices have existed on the social internet since the beginning. What's different now is that these platforms have two billion users. Or they have 400 hours of video content uploaded to them per minute per day. So the scale is really unprecedented. So is the need for workers to judicate material. A lot of the work is done by contractors. Are they working for Facebook or YouTube? Now or are they working for someone else. My team now sits in 11 offices around the world and these people are surface matter experts in a range of topics. I have somebody who was the director of West Point's counterterrorism research center. I have a woman whose doctorate is in extremist organizations and she's focused a lot on Europe. [ MUSIC ] I have two former human rights lawyers. I have somebody who is a teacher, somebody who is a rape crisis counselor. We are working everyday with members of our community operations team. That's the team of content reviewers, more than 7, 500 strong who are sitting around the world reviewing content 24 hours a day, seven days a week. We work with them to make sure that they understand the policies. We answer their questions. And if they're seeing something that doesn't look right, let's say they say to us, this doesn't count as bullying under our policies. But we're seeing a lot of it and it looks like bullying. Then we start a conversation about how we can refine that policy to make sure we're covering that new trend or that new conduct. These commercial content moderators, who are often in this decision making role, sort of has to set aside his or her own moral code, personal values, cultural context, working Elsewhere in the world under difficult productivity metrics where they might need to process thousands of images or videos a day. The human nuance that these moderators bring to the table is often disregarded in favor of clicking as fast as they can to get through the images. Well, there have been instances where something has been banned on the platform but then was deemed newsworthy and so was Allowed and where you've changed your decision. What's the process internally for those cases? Because we get so many reports every week, millions, we have to make sure that we have guidelines that can be implemented at scale. But we know that, at this scale, there will always be times when there is something that is close to the line Something that maybe technically doesn't violate our policies but when we look at it, we all agree, this is something that is hate speech. Something's are more easily done with the aid of computational methods and don't require as much human engagement. But some things are so challenging and nuance that human input is required to make decisions. How do we know that our particular symbol is being invoked? In a way to intimidate and perhaps not to show a historical case or record. These are the kinds of things that a moderator has to decide upon. Let's say that our technology says, this is a photo of a terror organizations flag or this is a photo of a hate organizations logo In either of these cases, that content might violate our policies if somebody is celebrating terrorism or a hate organization. But it might also be somebody who is raising about the organization, or a journalist who's posting a news story about something that this terrorist group has done. And in those cases we would wanna leave those post on the site. How do you think about measuring success for being able to enforce your policies? We certainly don't enforce these policies perfectly, because we have billions of posts coming on to the site, and we are not gonna become aware of And there are also times when content is reported to us and we simply get the decision wrong. These are real people on our community operations team who are reviewing this content and applying our standards and making these decisions. And we know that people make mistakes And even if our accuracy is higher than 99 %. At our scale, that still means we have thousands of mistakes everday. [ MUSIC ]

Will hiring thousands of moderators actually work?